Found this post, as well as another one that you posted in regards to this issue. We are having a similar issue starting in our 7.6 instance and persisting through to 9.1. Since we started in 3.6, we cannot proceed to 10.X until we upgrade our users to the "default" provider fron the "openaccess" provider, so not sure if this was resolved in a later version.
The reason for my email is that we are seeing the web worker process chew up 98% of the CPU and essentially bring the server to it's knees. we thought initially it was image related, but we hosted all images on disc, using a workaround that, if the image is seen with a default image source path, we re-route it. things were working well, but lately we have seen the issue rear it's head again, however now, the entire site wont load, for over 2 minutes sometimes.
We are on a Virtual hosted web server with 12 cores and 8 GB of installed memory.